This looks like an interesting cipher, but I would point out that the Python code does not do what the pdf file says it should. Specifically the permutation() function treats wrap around as a modulus function (e.g. leave on left hand end, rejoin on right) whereas the pdf (notably appendix C) treats it as a spiral - bounce back from the end we have reached - likely correct, given the name of the cipher. However the confusion is increased because the Python method has clearly been used to produce the sequence in Appendix C, whose letters don't actually align with the directions shown on the diagram.
There is a further ambiguity - it would be easy to read the spiral approach as follows : For letters ABCD… as we move left from D, do we count (below the line) D-C-B-A and then (above the line) A-B-C-D or do we count D-C-B-A-B-C-D (i.e. no double tap on A)? indeed the way the spiral cycles to the left potentially implies a triple-tap on A (below, left, above).
Worth being careful or waiting for confirmation of the approach. If the challenges have been produced with the Python code they won't align with the description, and vice-versa